`RDD.mapPartitions` should provide `iteratro` or `list` instead of concrete element

[adrian-dankiv]

RDD.mapPartitions in the mock works like RDD.map but shouldn't.

import dummy_spark
rdd = dummy_spark.RDD([1, 2, 3], None)
_ = rdd.map(print) # should print 1\n2\n3\n
1
2
3
# OK!
_ = rdd.mapPartitions(print) # should print [1, 2, 3] or <iterator object>
1
2
3
# FAIL!